Del Pezzo surfaces and representation theory (Q1005849)

The main object of the paper under review is a smooth del Pezzo surface \(X\) of degree \(d=9-r\) defined over a field \(k\) of characteristic zero. The surface is assumed \textit{split}, i.e. the Galois action on the set of exceptional curves is assumed trivial). After Manin discovered that the Picard lattice of \(X\) contains a root system \(R_r\) of rank \(r\) (\(r\geq 3\)), it has been a general feeling that this is a trace of some deeper relationship between rational surfaces and linear algebraic groups (or Lie algebras). An explicit form of such a relationship, at the level of torsors over del Pezzo surfaces, was suggested by Batyrev in 1990's. More precisely, his conjecture states that the universal torsor \(\mathcal T\) over \(X\) can be equivariantly (with respect to the action of the Néron--Severi torus of \(X\)) embedded, as an open set, into the affine cone \((G_r/P_r)_a\subset V(\omega_r)\) over the flag variety \(G_r/P_r\). Here \(G_r\) stands for the simply connected semisimple algebraic group of type \(R_r\); \(P_r\) is the maximal parabolic subgroup defined by the root \(\alpha_r\), which is to be erased from the Dynkin diagram when passing from \(R_r\) to \(R_{r-1}\); \(V(\omega_r)\) is the weight space of \(\omega_r\), the fundamental weight dual to the root \(\alpha_r\). In the paper under review, the authors prove this conjecture for \(d=2, 3, 4\). Note that the case \(d=5\) was settled earlier by the second author [Ann. Fac. Sci. Toulouse, VI. Sér., Math. 2, No. 3, 429--440 (1993; Zbl 0811.14020)]. The case \(d=4\) was settled by \textit{O.~N.~Popov} [Diplomarbeit, Univ. Tübingen, 2001], and the cases \(d=3\) and \(d=2\) by \textit{U.~Derenthal} [Adv. Math. 213, No. 2, 849--864 (2007; Zbl 1125.14020)], using a method of \textit{V.~V.~Batyrev} and \textit{O.~N.~Popov} [Prog. Math. 226, 85--103 (2004; Zbl 1075.14035)] based on considering the Cox ring of \(X\) (whose spectrum contains \(\mathcal T\) as an open subset), identifying its generators with weights, and comparing the relations with the equations of \(G_r/P_r\). The authors of the paper under review argue quite differently using the representation theory of Lie algebras. They proceed by induction, starting with the known case \(d=5\), and using (in the induction step) the fact that the pair \((R_r,\alpha_r)\) is a Hermitian symmetric pair for \(4\leq r\leq 7\) (i.e. the complementary nilpotent algebra to \(\text{Lie}(P_r)\) to \(\text{Lie}(G_r)\) is commutative). This allows them to interpret the blowing-up of a point on \(X\) in Lie-algebraic terms. Note that recently, using the same approach, the authors extended their result to the nonsplit case [\url{arXiv:0806.0089}, to appear in J. Inst. Math. Jussieu].